Fuzzy Based Model to Detect Patient’s Health Decline in Ambient

Assisted Living

Milene Santos Teixeira

, Vinícius Maran

, João Carlos D. Lima

, Iara Augustin

and Alencar Machado

Centro de Tecnologia, Universidade Federal de Santa Maria, Santa Maria - RS, Brazil

Coordenadoria Acadêmica, Universidade Federal de Santa Maria, Cachoeira do Sul - RS, Brazil

Colégio Politécnico, Universidade Federal de Santa Maria, Santa Maria - RS, Brazil

Keywords: Ambient Assisted Living, Health Decline Detection, Fuzzy Logic, Fuzzy Model.

Abstract: Detecting a decline in the health condition of a patient may still be considered a challenge in Ambient

Assisted Living (AAL) since the concept of ‘decline’ is vague and imprecise. In this context, Fuzzy Logic

comes as an excellent alternative for AAL systems. This paper presents a model based on Fuzzy logic

reasoning in order to identify a possible decline in the patient health condition. In order to achieve this goal,

the model considers relevant situations that may somehow impact the patient. To evaluate the model, a case

study was developed, showing that the developed model can simulate the human reasoning and be used in

an AAL system.

1 INTRODUCTION

Fuzzy logic (Bai; Wang, 2006) is present in our

daily lives. Some concepts used in everyday life,

such as tall, cold, young, among others, may be

vague and difficult to be clearly defined without

generating ambiguity. For example, a person may

say that a 28-year-old person is young, but another

may not agree with this point of view.

Fuzzy logic is an extension of the classic

Boolean logic that, by introducing the concept of

degree, makes it possible to present much more

precise results when dealing with uncertainty and

vagueness.

This logic aims at formalizing the human

reasoning in a much more natural way than it would

be done by just applying 0's or 1's, values known by

machines. Lately, this kind of reasoning has been

applied in complex systems (Marro et al., 2010)

such as expert systems and systems that deal with

artificial intelligence or human behaviour.

Ambient Assisted Living (AAL) (Pieper;

Antona; Cortés, 2011) is a field that, most of time,

has to interpret the human behavior. AAL applies

technology in order to assist elderly people in their

daily lives aiming at providing a safer environment

and, consequently, improve their life quality.

AAL also has as its goal to provide more

independence for the patient, making it possible for

the elderly person to live longer in his/her residence,

many times, without the need of the presence of a

caregiver.

However, despite all technology provided, it is

common that some patients present a natural decline

in their health condition, what should be perceived

by AAL systems. Nonetheless, 'detecting a decline'

can be seen as a very vague idea since it is hard to

define and describe what exactly means a decline. A

patient’s decline can be defined differently by two

different doctors, for example.

Additionally, another important aspect to be

noticed is the fact that the data obtained through

sensors may be imprecise and, many times,

incomplete. In this context, Fuzzy Logic can be

applied aiming to present a much more reasonable

solution, since its main goal is the computational

modeling of the human reasoning, which is

imprecise and vague (Marro et al., 2010).

This paper defines a model that considers

relevant situations for a patient in an AAL

environment in order to detect whether he is

presenting a decline in his health condition.

Considering the fact that detecting a situation

(health decline, in our case) presents a considerable

level of uncertainty summed to the feature of

Teixeira, M., Maran, V., Lima, J., Augustin, I. and Machado, A.

Fuzzy Based Model to Detect Patient’s Health Decline in Ambient Assisted Living.

DOI: 10.5220/0006368806590666

In Proceedings of the 19th International Conference on Enterprise Information Systems (ICEIS 2017) - Volume 1, pages 659-666

ISBN: 978-989-758-247-9

659

human-like language of the Fuzzy logic, this work

will facilitate caregivers’ work in an AAL. Our

model aims at achieving a result closer to what

would be achieved by human reasoning and can be

implemented allied with other already published

reasoning models in AAL. In addition, since an

AAL environment presents multiple applications and

sensors, a new application making use of this model

can be easily introduced to the environment.

This paper is structured as follows: Section 2

presents a background of the concepts used in this

work. In Section 3, related works are presented. The

developed model is presented in Section 4 followed

by Section 5 that presents a case study

demonstrating the use of this model in a case study.

Finally, in Section 6, the conclusions and future

works are presented.

2 BACKGOUND

In order to make the reader more familiar with the

model developed in this paper, this section explains

the concept of Fuzzy Logic and the three steps of its

process (fuzzification, inference, deffuzification).

We also present the main concepts related to AAL

systems and its main goals.

2.1 Fuzzy Logic

The concept of Fuzzy logic was introduced in 1965

by Lotfi Zadé. According to Zadé (Zadé, 1965),

terms such 'fast' and 'hot' can be created and

implemented only by human beings. In other words,

computers are not capable of reasoning on such

terms, because they only interpret the meaning of 0

and 1 (Bai; Wang, 2006). He also highlights that

such way of expressivity exerts an important role on

the logic and human reasoning.

In this way, Fuzzy logic introduces the concept

of degree instead of getting limited to only 'true' or

'false'. In Fuzzy logic everything can present a

degree that is represented by a word or by a numeric

value (usually the interval 0...1 is used), indicating

the degree of veracity of the information.

The Fuzzy processing is a crisp-Fuzzy-crisp

process (Bai; Wang, 2006). It means that (i) input

values are crisp (classical logic), (ii) in order to be

processed, these values are converted to Fuzzy, and

(iii) the result is converted back to crisp.

The most well-known method for achieving the

Fuzzy inference process, is the Mandami method

(Marro et al., 2010). According to Mandami, the

three necessary steps to implement Fuzzy logic in an

application consist of: fuzzification, process of

Fuzzy inference and defuzzification.

2.1.1 Fuzzification

In order to make it possible for machines to process

vague information, it is necessary to convert the

input and output data, so far crisp (numeric), to

linguistic variables with Fuzzy components. To

achieve this goal in the Fuzzification process, first,

the membership functions (μ(x)) are defined to every

possible input or output variable.

Membership functions (George; Bo, 2008)

indicate the degree that an element belongs to a

given set. For example, given an input variable

'temperature', three membership functions are

defined: 'cold' (0-20), 'normal' (10-30) and 'warm'

(20-40). After this definition, the input values

(usually obtained through sensors, other systems,

etc.) are calculated identifying what is known as the

membership degree to each membership function

defined. Supposing that the input value for

temperature is 12°, the membership function will

identify a membership degree greater than 0 for

'cold' and 'normal', but 0 for 'warm'.

2.1.2 Fuzzy Inference Process

In this step, the membership degrees are combined

with inference rules (Bai; Wang, 2006) in order to

obtain a Fuzzy output value. The inference rules are

generated based on the human knowledge and

experience over the application domain. These rules

are known as if-then rules, representing what action

should be taken or what information is obtained

based on the input value. An example related to

temperature is given below:

IF temperature IS high

THEN reduceSpeed.

(Vieira, 1999) highlights that there is not a

mathematical formulation to generate these rules.

They are usually defined by an expert and must be

very intuitive (Bai; Wang, 2006) (Vieira, 1999),

presenting as aspect to be easy to understand and

comprehend when read by humans.

2.1.3 Defuzzification

This step consists in an inverse process in relation to

Fuzzification. In other words, the result obtained is

converted from linguistic back to crisp (numeric)

and, then, can be applied in an application or system.

There are different methods available that can be

ICEIS 2017 - 19th International Conference on Enterprise Information Systems

660

applied in this process. However, Center of Gravity

(COG) (Dernoncourt, 2011) is the most commonly

used method.

Due to the characteristic of the use of natural

language, Fuzzy logic suits well in technologies that

deal with human reasoning and behaviour. Expert

systems and Ambient Intelligence are examples of

technologies that present meaningful research and

works implementing Fuzzy logic (Marro et al.,

2010). Ambient Assisted Living is also a field that,

many times, has to interpret vague and imprecise

concepts and, also, try to reproduce the human

reasoning. This field is better described in the next

section.

2.2 Ambient Assisted Living (AAL)

Ambient Assisted Living (AAL) is a field within

Ambient Intelligence (AmI) that focuses on the use

of ubiquitous or pervasive technology in elderly’s

residences (Aarts; Wichert, 2009). This field has

been widely explored in several researches in the

field of Computer Science (Pieper; Antona; Cortés,

2011).

AAL aims at creating a safe environment able to

improve elderly’s autonomy and to assist them in

daily activities, making it possible at the same time,

to preserve their independence (Sun et al., 2009).

One of its main goals is to extend the time that the

elder will be able to continue living in his/her

residence without the need of the care of a third part.

Some examples of technologies (Rashidi;

Mihailidis, 2013) in AAL are: applications capable

of alerting the patient to the correct use of his

medicine; sensors that detect possible falls; and

robots able to help with simple routine tasks, such as

sweeping the floor or washing dishes.

In other words, AAL covers concepts, products

and services that connect new technologies to the

patient’s own environment, being it possible to be

used in the prevention, cure and improvement of his

health condition as well as of his wellbeing.

3 RELATED WORK

It is possible to find many works in the literature that

aim at contributing to detect and/or avoid unwanted

situations in AAL. Many of these works report the

inconvenience of devices that require some attention

of the user to operate.

In (Storf et al., 2009), for example, an approach

was described for detecting situations in an AAL

environment in order to act in a proactive way and

avoid emergencies. Their approach focused on the

use of non-obtrusive devices and this was achieved

by the use of information obtained from sensors in

the environment and by the analysis of daily and

historical data of the patient.

Based on the fact that many times an undesired

situation may occur as a result of some other

previous situations or actions mistaken by the

patient, (Machado et al., 2017) presented an

approach that makes use of Bayesian Networks and

aims at an early detection of these undesired

situations in order to avoid them. With this, this

work presented the importance of mechanisms that

act not only reactively, but also proactively in AAL.

When it comes to the use of Fuzzy Logic in

AAL, we can give as an example the system

proposed by (Nefti; Manzoor; Manzoor, 2010).

Their work consisted of a “multi-agent system based

application used to assess the risk level in different

situations to patient”, which goal is to monitor

patients suffering from dementia. Fuzzy logic was

applied to predict the risk assessment, as it can

simulate human-like decisions to determine the best

course of actions to be taken.

By the analysis of works like the mentioned

ones, we can see the importance of detecting

situations in order to avoid future unwanted ones.

Another aspect to be considered is, since Fuzzy

logic makes use of a very human-like language, we

realize that Fuzzy logic makes it possible to users

who are not information systems experts to

reconfigure the system. This would not be achieved

in a trivial way by the use of Bayesian Networks, for

example.

(Walley, 1996) compares some measures that are

used when dealing with uncertainty in expert

systems. Since he describes and compare general

features on Fuzzy Logic and Bayesyan Probabilities

(among others), his comparison can also be applied

to the context of this paper.

Walley highlights the contribution of the Fuzzy

Logic to expert systems due to the design of natural

language reasoning and the introduction of

possibility measures models. According to him,

Fuzzy logic has the advantage of handling well the

aspect ‘imprecision’, since it is one of its main goals.

In addition, it also addresses well the aspect

'assessment', since it makes it possible to the users of

the system to use natural language when describing

the uncertainty on the domain.

One advantage of Bayesian probabilities is that

they guarantee consistency in their rules, what may

be different in Fuzzy Logic (Walley, 1996).

However, they do not handle well aspects such as

uncertainty in natural language and incomplete

Fuzzy Based Model to Detect Patient’s Health Decline in Ambient Assisted Living

661

information. Also, Bayesian probabilities are not

satisfactory when it comes to modeling vague terms

or vague probabilities as, for example, 'Mary is

probably young'. Bayesians demand precise

probability models and lack in the aspect

'imprecision' being, therefore, not suitable for our

work.

The model presented in this paper does not

intend to replace neither put in question any other

work previously published. Instead, it creates the

possibility of joining some different existing models

in order to achieve a more specific goal in AAL (to

detect a decline). Such feature was not attended by

any of the previously mentioned researches and was

even being ignored by the execution of actions that

avoid undesired situations but do not identify a

cognitive decline.

The next section presents the model developed in

this work.

4 FUZZY BASED MODEL FOR

AAL

Within the set of vague concepts that cannot be

simply identified with a ‘true or false’ value without

generating ambiguity or the need for further

explanation, is the concept of ‘health decline’. In

order to make it possible to a system to achieve a

result closer to a human being reasoning in this

vague context, we developed a model based on

Fuzzy logic to detect a possible health decline of a

patient in an AAL environment.

The model considers that a decline degree can be

achieved after the defuzzification of the impact

degree obtained from different situations considered

relevant for the patient. More specifically, we

establish that a health decline is influenced by the

occurrence of situations that affect negatively the

patient, the impact that they have on his wellbeing,

and it is aggravated by the recurrence of these

situations within a certain period of time (Figure 1).

For the concept of situation, we adopted the

concept used in (Machado et al., 2017) that

describes a situation as a set of active entities and

the interactions between them in a frame of time. In

other words, a situation has a start and an end time,

is composed of entities, their attributes and the

relations between these entities. In the AAL context,

some examples of situations can be: Patient felt,

Patient had a heart attack and Patient forgot taking

a medicine.

Figure 1: Health decline model. T: time interval, S:

situation, I: impact degree, NO: number of occurrences.

It is important to point out that every patient

may present different relevant situations, identified

in the model as linguistic variables to be considered.

For example, for some patients it is important to

consider possible falls, however, for a patient in a

wheelchair this information may have no relevance.

This work does not detect these situations, our

focus is on the detection of a possible health decline.

Therefore, in order to identify the relevant situations,

some related works can be applied as, for example,

the framework presented in (Machado et al., 2016).

This framework makes use of a Multi-Entity

Bayesian Network (MEBN) and presents an

ontology network in order to predict unwanted

situations in smart environments.

Our Fuzzy based model (c) depends on a

database (a) containing data about the patient

(usually obtained from sensors or other systems) and

a rule base (b) (may be specified by an expert in the

domain or by machine learning). After the fuzzy

reasoning, the information obtained is available to

be used by any external system (d). Figure 2

illustrates the model (c) and its external connections

(a, b and d).

Figure 2: Fuzzy based model and its external connections.

The model developed is composed of 3 stages

(Figure 3). These stages are described in the

sequence:

4.1 Stage 1

In the first stage, all relevant input linguistic

variables (a.1, b.1, … , n.1) should be detected

presenting the corresponding membership functions.

These variables correspond to situations faced by the

patient that present some risk to his wellbeing

ICEIS 2017 - 19th International Conference on Enterprise Information Systems

662

Figure 3: Fuzzy model.

impacting in a health decline. For each linguistic

variable, a membership degree (μ(S)=x) must be

identified, which is the input used to achieve the

impact degree in the fuzzy controller.

As mentioned before, the situations may be

different for each patient and can be obtained from

different sources. Besides the framework cited, other

examples of sources are experts and ontologies.

Nonetheless, the membership degree should be

defined by an expert or it can be the output

originated from the processing of previous linguistic

variables in a Fuzzy controller.

For example, let us suppose that one of the

variables that is being considered is the level of

forgetfulness of a patient over his medicines

(level_forget: {low, moderate, high}). Detecting this

level, may not be possible without, among other

relevant coefficients, analysing the importance of the

medicine (medicine_relevance:{notMuch, moderate,

important}) and for how long the medicine was

forgotten (medicine_forgot: {shortPeriod,

moderatePeriod, longPeriod}. These data could

have been pre-processed by a Fuzzy controller

being, therefore, level_forget value the

defuzzificated output of it.

Associated to each linguistic variable defined,

there must be a second linguistic variable (a.2) that

defines the number of occurrences

(num_occurrences) of this situation in some interval

of time defined according to the need. The

membership functions defined for num_occurrences

in this model are 5, being labeled as: never, few,

moderate, many, always. Its bounds vary according

to the interval of time defined. This leads us to

conclude that there must exist a database containing

historical data obtained from sensors or applications

available for this patient.

4.2 Stage 2

Impact degree (impact_degree) is also defined as a

linguistic variable (output), being its membership

degree (μ(I)=x) the output resulting from the

execution of the inference rules applied to the

membership functions of the linguistic variables

previously defined. This process must happen in

each input linguistic variable identified.

The possible source for retrieval of the inference

rules could be: an expert; machine learning; a base

of rules previously defined in the system; another

system or model that makes use of methods as the

one presented by (De Campos; Moral, 1993).

The membership functions for the variable

impact_degree are labeled as: low (0-4), moderate

(1-9), high (6-10).

Again, it is important to remember that the

linguistic variable related to the number of

occurrences must be always considered in an

inference rule. For example:

IF level_forget IS high

AND num_occurrences IS many

THEN risk IS HIGH

IF level_falling IS moderate

AND num_occurrences IS moderate

THEN risk IS HIGH

4.3 Stage 3

After having the membership degree values for

impact_degree, it is finally possible to figure out

whether or not a decline in the patient's health

condition is happening. Basically, in this stage, after

aggregating the membership degrees for

impact_degree and identifying the resulting fuzzy

set for decline_level (low, moderate, high), there

will happen the process of defuzzification. Further

studies should be realized in order to identify the

most recommended defuzzification technique to be

applied in this model. Meanwhile, we recommend to

use COG, the most commonly used one.

We consider that for a result with ‘0’ value

means that there is absolutely no decline in the

patient situation, however the ‘10’ value means that

there is a 'complete' decline and some providence

should be taken as fast as possible. For any different

value in the interval 0-10, the application using this

model should decide how to proceed.

Fuzzy Based Model to Detect Patient’s Health Decline in Ambient Assisted Living

663

5 CASE STUDY

In order to demonstrate the use of the presented

model, we defined a fictitious scenario of a patient

living in an AAL environment. The Fuzzy controller

tool available in Matlab (Fuzzy Logic toolbox) was

used to process the data.

Let us consider the following scenario: “Mr.

Miller is a 77 year old retired citizen who lives in an

AAL residence and has some aging-associated

diseases such as memory disorders, hypertension

and Parkinson's disease (what can increase the risk

of falls). Because of his condition, he takes different

kinds of controlled medicines. Each medicine

presents a different impact and relevance in his

treatment in case of omission. Considering that Mr.

Miller's current health situation is defined as stable,

he does not need the constant presence of his

caregiver. However, in the AAL residence, there is a

middleware (Machado et al., 2017) that uses the

given Fuzzy based model and constantly monitors

his health situation in order to identify a decline and

possible need of a full-time caregiver.”

Considering the scenario described and the

model defined in the previous section, the following

activities must be identified in order to implement

the model:

5.1 Identification of Possible Risky

Situations

With the help of Mr. Miller's physician and a

database containing a history of his situation from

the last two months (falls, medicine's usage,

hospitalizations, among others), two situations are

identified as possibly offering some risk to his

treatment: falls (situation A) and forgetfulness of

medicine usage (situation B). In this context, the

two linguistic variables defined are:

(a) Falls_Level: Represents a history of falls of

the patient. The membership functions are labelled

as light_fall (when there was no injury and the

patient straightened up by himself; interval: 0-3.5),

moderate_fall (the patient needed help to stand up

and there may be an injury; interval: 2.5-5),

heavy_fall (the patient was injured and possibly

hospitalized; interval: 4-10);

(b) Forgetfulness_Level: represents a historical

of the patient's forgetfulness of the usage of his

medicines. The membership functions are labelled as

low (0-3.5), moderate (3-6), high (5-10).

Considering the registers contained in the

database, we suppose that both situations are first

processed in a Fuzzy controller in order to achieve a

value that will be used as the input value in our

model. In other words, the membership degrees for

both linguistic variables are the result of the output

of another Fuzzy reasoning.

In situation A, for example, all registers of falls

were analysed considering their gravity (resulted in

injury or not), whether someone had to be contacted,

etc. Then these registers were processed in a Fuzzy

controller and the value 4.2 was obtained. Similarly,

the registers for medicines usage were analysed

considering whether the medicine was taken late or

it was not taken, the relevance of the medicine, its

acceptable delay, among other factors.

Table 1 shows the input values identified, as well

as the membership degree for the variable

num_occurrences of each of these variables.

Table 1: Input values.

Situation Value Number of

occurrences

A 4.2 6

B 3.8 18

The number of occurrences considers all

registers, independently of their level/degree. As the

period considered is two months, the bounds for the

membership functions in num_occurrences are

defined in days as follows: never (0-5), few (4-15),

moderate (10-25), many (18-55), always (53-60).

Finally, we have the four input variables (Figure

4) that will be used in our Fuzzy controller, which

will have as output 'impactA' for the impact level

identified.

Figure 4: Input linguist variables. (a) fallsLevel, (b)

forgetfullnessLevel, (c) numOccurrencesA, (d)

numOccurrencesB.

5.2 Identify the Membership Degree of

Impact for Each Defined Linguistic

Variable

To define the inference rules, it is required a high

level of knowledge about the situation in order to get

a good approximation to human reasoning. In this

ICEIS 2017 - 19th International Conference on Enterprise Information Systems

664

case, we consider that the rules were derived

according to the knowledge of Mr. Miller's

physician. The inference rules are illustrated in the

matrices in Figure 5.

Figure 5: Inference rules matrices.

Based on these matrices, it is possible to describe

linguistically the inference rules as, for example:

IF falls_level IS light

AND num_occurrencesA IS always

THEN impact IS high

IF forgetfullness_level IS high

AND num_occurrencesB IS few

THEN impact IS moderate.

Notice that the impact value increases (is

aggravated) as the number of occurrences also

increases.

5.3 Identifying Possible Decline

By applying the inference rules to our input

variables, impact membership degrees are obtained

and then aggregated in order to identify the resulting

Fuzzy set (Figure 6(d)).

Then, the defuzzification method COG is applied

and we, finally, have a result that identifies a

possible decline in the health condition of Mr. Miller

(Figure 6(b)).The result obtained in this case study

was 5.04, what represents a significant decline in the

health situation of the patient. Supposing that an

application that makes use of this model considers

that for any value above 4 means that the patient

needs the attention of a caregiver, this is the moment

that the application would alert Mr. Miller's health

provider. A physician analysing the data provided

summed to his knowledge about the context, would

probably achieve the same conclusion – there is a

significant decline. In this way, we achieve the goal

for this model - the result obtained simulates the

human reasoning using vague concepts.

Another advantage of the use of a system using

this model is the possibility of, when a decline is

identified, the caregiver/doctor can verify what is

influencing it in an easy to understand language. For

example, if the system makes it available the

inference rules and a graphic view of the inference,

the person analysing it will be able to identify the

linguistic variables (falls, medicine forgetfulness,

among others) and easily comprehend it. With that,

the routine or environment of the patient can suffer

adaptations in order to be improved.

Summarizing, the application of the model in the

case study made it possible to determine the health

decline of the patient involved in the scenario by

using vague concepts. In this way, the use of this

model allied with other reasoning models for AAL

(Maran et al., 2015) could be efficiently used by

AAL systems.

6 CONCLUSIONS

Ambient Assisted Living (AAL) is a field that deals

with people, their actions, behaviour, and even

Figure 6: Fuzzy inference. (a) crisp input values, (b) crisp final result, (c) Fuzzy rules applied, (d) Fuzzy resulting set.

Fuzzy Based Model to Detect Patient’s Health Decline in Ambient Assisted Living

665

emotions. None of these aspects is very precise

being it, many times, hard to define a yes or no,

black or white. Considering that detecting a decline

in the health situation of a patient in an AAL may be

vague and difficult, this paper presented a model that

makes use of Fuzzy Logic to achieve this goal.

To detect a health decline, our model considers

as input values daily situations that are faced by the

patient and may offer some risk to his wellbeing.

The impact of each situation is processed in a Fuzzy

controller and, finally, a value for the decline is

obtained. In order to better explain our model, we

presented a case study with a fictitious scenario.

We are aware that the model presented in this

paper is not the only possible way to detect a decline

in the health situation of a patient. Many other

methods can be applied, however, one of our goals

in this work is, through the use of Fuzzy logic (since

it deals with vagueness, uncertainty and aims to

reproduce human decisions), to achieve a result

much more close to the reality being it similar to a

result that could have been obtained if the situation

of the patient was being analysed by a human being

(expert, physician, among others) and not by a

computer limited to 0 and 1s.

As future work, we aim to elaborate an approach

that identifies automatically the situations that may

offer some risk to the patient generating, also

automatically, the membership degree to be used as

input in this model. We also aim to apply the model

in a system with real data in order to develop further

studies to determine the accuracy of the model

developed.

REFERENCES

Aarts, E. and Wichert, R., 2009. Ambient intelligence. In

Technology Guide (pp. 244-249). Springer Berlin

Heidelberg.

Bai, Y. and Wang, D., 2006. Fundamentals of fuzzy logic

control—fuzzy sets, fuzzy rules and defuzzifications.

In Advanced Fuzzy Logic Technologies in Industrial

Applications (pp. 17-36). Springer London.

De Campos, L.M. and Moral, S., 1993. Learning rules for

a fuzzy inference model. Fuzzy Sets and Systems,

59(3), pp.247-257.

Dernoncourt, F., 2011. Introduction to fuzzy logic.

George J, K.L.I.R. and Bo, Y., 2008. Fuzzy sets and fuzzy

logic, theory and applications. -.

Machado, A., Maran V., Augustin, I., Lima, J.C., Wives,

L.K. and de Oliveira, J.P.M., 2016. Reasoning on

Uncertainty in Smart Environments. In Proceedings of

the 18th International Conference on Enterprise

Information Systems - Volume 2: ICEIS, 240-250,

Rome, Italy.

Machado, A., Maran, V., Augustin, I., Wives, L.K. and de

Oliveira, J.P.M., 2017. Reactive, proactive, and

extensible situation-awareness in ambient assisted

living. Expert Systems with Applications, 76, pp.21-35.

Maran, V., Machado, A., Augustin, I., Wives, L.K., and de

Oliveira, J.P.M., 2015. Proactive Domain Data

Querying based on Context Information in Ambient

Assisted Living Environments.

Marro, A.A., de Carvalho Souza, A.M., de Sousa

Cavalcante, E.R., Bezerra, G.S. and de Oliveira

Nunes, R., 2010. Lógica fuzzy: conceitos e aplicações.

Natal: Universidade Federal do Rio Grande do Norte

(UFRN).

Nefti, S., Manzoor, U. and Manzoor, S., 2010, June.

Cognitive agent based intelligent warning system to

monitor patients suffering from dementia using

ambient assisted living. In Information Society (i-

Society), 2010 International Conference on (pp. 92-

97). IEEE.

Pieper, M., Antona, M. and Cortés, U., 2011. Ambient

assisted living. Ercim News, (87). Available in:

http://ercim-

news.ercim.eu/images/stories/EN87/EN87-web.pdf.

Accessed in: September 15, 2016.

Rashidi, P. and Mihailidis, A., 2013. A survey on ambient-

assisted living tools for older adults. IEEE journal of

biomedical and health informatics, 17(3), pp.579-590.

Storf, H., Kleinberger, T., Becker, M., Schmitt, M.,

Bomarius, F. and Prueckner, S., 2009, November. An

event-driven approach to activity recognition in

ambient assisted living. In European Conference on

Ambient Intelligence (pp. 123-132). Springer Berlin

Heidelberg.

Sun, H., De Florio, V., Gui, N. and Blondia, C., 2009,

April. Promises and challenges of ambient assisted

living systems. In Information Technology: New

Generations, 2009. ITNG'09. Sixth International

Conference on (pp. 1201-1207). Ieee.

Vieira, R.D.S., 1999. Protótipo de um sistema de

monitoramento remoto inteligente.

Walley, P., 1996. Measures of uncertainty in expert

systems. Artificial intelligence, 83(1), pp.1-58.

Zadé L. A., 1965. Fuzzy Sets. Intl J. Information Control

8:338-353.

ICEIS 2017 - 19th International Conference on Enterprise Information Systems

666